Cloud assisted machine learning

Abstract

A method for training an analytics engine hosted by an edge server device is provided. The method includes determining a classification for data in an analytics engine hosted by an edge server and computing a confidence level for the classification. The confidence level is compared to a threshold. The data is sent to a cloud server if the confidence level is less than the threshold. A reclassification is received from the cloud server and the analytics engine is trained based, at least in part, on the data and the reclassification.

What is claimed is: 
     
       1. An edge server comprising:
 a processor; and 
 memory to store instructions to direct the processor to:
 analyze data from an Internet-of-things (IoT) sensor to predict an event based on the data; 
 identify, using a prediction model, a first weighted average confidence level for a first analysis result relating to the event; 
 separate the analyzed data based on subject categories; 
 assign a confidence level to the analyzed data for each subject category of the subject categories; 
 compute the first weighted average confidence level based on the assigned confidence levels; 
 send the data to a cloud server for processing if the first weighted average confidence level is less than a first threshold; 
 receive a second analysis result from the cloud server, wherein the second analysis result has a second weighted average confidence level greater than a second threshold; and 
 train the prediction model based on the second analysis result. 
 
 
     
     
       2. The edge server of  claim 1 , wherein the edge server is interconnected with a plurality of Internet-of-things (IoT) devices. 
     
     
       3. The edge server of  claim 1 , wherein the memory includes instructions to direct the processor to:
 generate a unique identification code for the data; and 
 send the unique identification code to the cloud server. 
 
     
     
       4. The edge server of  claim 1 , wherein the memory includes instructions to direct the processor to:
 store the data sent to the cloud server in a data store; and 
 associate the data in the data store with a unique identification code. 
 
     
     
       5. The edge server of  claim 1 , wherein the memory includes instructions to direct the processor to:
 access the data in a data store using a unique identification code returned from the cloud server. 
 
     
     
       6. The edge server of  claim 1 , wherein the first analysis result, the second analysis result, or both the first analysis result and the second analysis result includes a classification of the data. 
     
     
       7. The edge server of  claim 1 , wherein data is discarded if the first weighted average confidence level is less than a threshold. 
     
     
       8. The edge server of  claim 1 , comprising a display to display content based on the first analysis result or the second analysis result. 
     
     
       9. The edge server of  claim 1 , wherein the cloud server is to implement cloud analytics. 
     
     
       10. The edge server of  claim 9 , wherein the cloud analytics is to generate a reclassification confidence level for the second analysis result. 
     
     
       11. The edge server of  claim 10 , wherein the cloud analytics is to discard the data if the reclassification confidence level is less than a cloud threshold. 
     
     
       12. The edge server of  claim 1 , comprising an anonymous video analyzer (AVA). 
     
     
       13. The edge server of  claim 1 , wherein the first threshold is the same as the second threshold. 
     
     
       14. The edge server of  claim 1 , wherein the second threshold is higher than the first threshold. 
     
     
       15. The edge server of  claim 1 , wherein the data comprises images of an object approaching a controlled roadway intersection, wherein the event comprises a time of arrival of the object at the controlled roadway intersection, and wherein the first analysis result comprises a prediction of when the object will arrive at the controlled roadway intersection. 
     
     
       16. The edge server of  claim 1 , wherein the instructions stored in memory are implemented as a micro-service, and the micro-service is distributed to a remote device communicatively coupled to the edge server. 
     
     
       17. The edge server of  claim 16 , wherein the remote device distributes the micro-service to another remote device. 
     
     
       18. The edge server of  claim 16 , wherein the remote device removes the micro-service after being used. 
     
     
       19. A method comprising:
 analyzing data from an Internet-of-things (IoT) sensor to predict an event based on the data; 
 identifying, using a prediction model, a first weighted average confidence level for a first analysis result relating to the event; 
 separating the analyzed data based on subject categories; 
 assigning a confidence level to the analyzed data for each subject category of the subject categories; 
 computing the first weighted average confidence level based on the assigned confidence levels; 
 sending the data to a cloud server for processing if the first weighted average confidence level is less than a first threshold; 
 receiving a second analysis result from the cloud server, wherein the second analysis result has a second weighted average confidence level greater than a second threshold; and 
 training the prediction model based on the second analysis result. 
 
     
     
       20. The method of  claim 19 , wherein the first analysis result, the second analysis result, or both the first analysis result and the second analysis result includes a classification of the data. 
     
     
       21. The method of  claim 19 , wherein the cloud server is to implement cloud analytics to generate a reclassification confidence level for the second analysis result. 
     
     
       22. The method of  claim 21 , wherein the cloud analytics is to discard the data if the reclassification confidence level is less than a cloud threshold. 
     
     
       23. The method of  claim 19 , wherein the second threshold is higher than the first threshold. 
     
     
       24. At least one non-transitory machine-readable medium including instructions, which when executed by a machine, cause the machine to:
 analyze data from an Internet-of-things (IoT) sensor to predict an event based on the data; 
 identify, using a prediction model, a first weighted average confidence level for a first analysis result relating to the event; 
 separate the analyzed data based on subject categories; 
 assign a confidence level to the analyzed data for each subject category of the subject categories; 
 compute the first weighted average confidence level based on the assigned confidence levels; 
 send the data to a cloud server for processing if the first weighted average confidence level is less than a first threshold; 
 receive a second analysis result from the cloud server, wherein the second analysis result has a second weighted average confidence level greater than a second threshold; and 
 train the prediction model based on the second analysis result. 
 
     
     
       25. The at least one non-transitory machine-readable medium of  claim 24 , wherein the second threshold is higher than the first threshold.